A warehouse management system and method

Abstract

The application provides a warehouse management system and method, relates to the technical field of warehouse management, and aims to improve the efficiency and accuracy of warehouse equipment taking, and enhance the safety and reliability of the warehouse storage and taking process. The system comprises an access control device, a plurality of equipment storage positions and a central control platform. The central control platform is used for: generating a pass authorization instruction and a list of equipment to be taken after the right of the equipment taking application is audited and passed; controlling the access control device to open the access control after the biological feature verification of the user's face image and the pass authorization instruction is passed; calculating a target path from the access control position to the target storage position in the list of equipment to be taken, sequentially lighting the display units of the equipment storage positions on the target path according to the time sequence delay, forming a dynamic light guide; comparing the read real-time label code with the target equipment code in the list of equipment to be taken, and controlling the electronic lock of the target storage position to execute the equipment release operation or trigger an abnormal prompt based on the comparison result.

Description

Technical Field

[0001] This application relates to the field of warehouse management technology, and in particular to a warehouse management system and method. Background Technology

[0002] With the rapid development of logistics warehousing, power operation and maintenance, and military field operations, the requirements for warehouse security management are increasing. Traditional skid-mounted warehouses, in particular, are widely used for equipment storage and material turnover due to their convenient transportation, rapid assembly, and strong adaptability. However, traditional skid-mounted warehouses face significant technical bottlenecks in intelligent management: access control relies heavily on IC cards or mechanical keys, which are difficult to defend against forgery attacks; environmental monitoring mostly only has single-threshold alarm functions, resulting in large fluctuations in temperature and humidity within the warehouse, failing to meet the storage requirements of precision instruments; equipment positioning and guidance rely on manual searching or simple label indications, leading to low retrieval efficiency and difficulty in real-time detection of misplacement or displacement. In short, related skid-mounted warehouses suffer from low levels of intelligence, insufficient security, and poor retrieval efficiency in areas such as access control, environmental regulation, and retrieval guidance. Therefore, there is an urgent need for a warehouse management system and method that can achieve intelligent closed-loop management throughout the entire process, improving both security and retrieval efficiency. Summary of the Invention

[0003] The purpose of this application is to provide a warehouse management system and method, which aims to improve the efficiency and accuracy of warehouse equipment retrieval, while enhancing the safety and reliability of the warehouse storage and retrieval process.

[0004] To achieve the above objectives, the embodiments of this application provide the following technical solutions: Firstly, this application provides a warehouse management system, comprising: an access control device located at the warehouse entrance for capturing user facial images; multiple device storage slots located inside the warehouse, each slot for storing devices to be retrieved, each slot including a display unit, a reading device, and an electronic lock; the display unit for displaying guidance information; the reading device for reading the tag code of the stored device; and the electronic lock for locking or releasing the stored devices to be retrieved; and a central control platform connected to the access control device and the multiple device storage slots, configured to: respond to device retrieval requests submitted by user terminals, perform authorization verification, and, within authorized limits... When the approval result is passed, an access authorization instruction and a list of devices to be retrieved are generated. Biometric verification is performed based on the collected user facial image and access authorization instruction. After the biometric verification result is passed, the access control device is controlled to open the access control. According to the storage coordinates of the target storage location in the list of devices to be retrieved, the target path from the access control location to the target storage location is calculated. The display units of each device storage location on the target path are lit up in sequence according to the time delay to form dynamic light guidance. The real-time tag code read by the reading device of the target storage location is compared with the target device code in the list of devices to be retrieved. Based on the comparison result, the electronic lock of the target storage location is controlled to perform the device release operation or trigger an abnormal prompt.

[0005] Beneficial Effects: This warehouse management system includes access control devices, multiple equipment storage locations, and a central control platform. The central control platform first reviews user-submitted equipment retrieval requests, generating access authorization instructions and a list of equipment to be retrieved. This ensures that only approved, legitimate requests proceed to the next stage, preventing irrelevant requests from directly accessing the warehouse. Next, biometric verification is performed based on facial images captured by the access control devices. Access is only opened after successful verification, ensuring the actual person present matches the authorized identity and preventing identity theft. Once inside the warehouse, the optimal path from the access control location to the target storage location is calculated based on the storage coordinates of the target storage location in the list of equipment to be retrieved. The storage locations of each device along the path are then illuminated sequentially according to time. The display unit at each location provides dynamic lighting guidance, allowing users to quickly and accurately reach their target location along the light strip without needing to memorize or search. This significantly improves retrieval efficiency and reduces the possibility of picking the wrong shelf. When a user retrieves equipment from the target storage location, the system reads the equipment's real-time tag code and compares it with the target equipment code in the inventory list. If they match, the electronic lock is activated to release the equipment; otherwise, an error message is triggered. This creates a closed-loop verification process, effectively preventing equipment from being taken by mistake, ensuring accurate inventory, and providing immediate alerts to personnel, thus enhancing the security and reliability of equipment storage and retrieval within the warehouse.

[0006] In some embodiments, the central control platform is configured to compare the real-time tag code read by the reading device of the target storage location with the target device code in the list of devices to be retrieved, and control the target storage location to perform a device release operation or trigger an abnormal prompt based on the comparison result, including: if the comparison result of the real-time tag code and the target device code is consistent, controlling the target storage location to open the release and recording a successful device retrieval event; if the comparison result of the real-time tag code and the target device code is inconsistent, determining the alarm level based on the feature similarity between the device in the target storage location and the target device corresponding to the target device code; and controlling the alarm volume and the alarm information push range based on the alarm level.

[0007] In some embodiments, the central control platform is configured to calculate a target path from the access control location to the target storage location based on the storage coordinates of the target storage location in the list of devices to be retrieved. This includes: determining the shortest walking path based on a two-dimensional electronic map of the warehouse, with the access control location as the starting node and the target storage location as the ending node; the two-dimensional electronic map of the warehouse is a pre-stored floor plan of the warehouse interior, which is marked with the physical coordinates of each device storage location and the walking distance between each adjacent node; a preset path algorithm is used to search the two-dimensional electronic map from the starting node, expanding the search level by level to the ending node, and selecting the path with the shortest total walking distance; the target path is decomposed into a sequence of nodes, and the display units of the corresponding device storage locations of each node are lit sequentially from the starting node to the ending node according to a preset time interval.

[0008] In some embodiments, the central control platform is configured to perform biometric verification based on the collected user face image and access authorization command, and control the access control device to open the access control after the biometric verification result is passed. This includes: extracting the facial feature vector of the collected face image, comparing it with the pre-stored standard feature template in the access authorization command to determine a comprehensive judgment score; performing liveness detection based on the user's action data to determine the liveness detection result; and determining that the biometric verification result is passed and the access control device is controlled to open the access control when the comprehensive judgment score is greater than a preset threshold and the liveness detection result meets preset conditions.

[0009] In some embodiments, the central control platform is configured to extract facial feature vectors from the acquired face images, compare them with the pre-stored standard feature templates in the access authorization instructions, and determine a comprehensive judgment score, including: determining a first similarity based on the Euclidean distance between the facial feature vectors and the standard feature templates; determining a second similarity based on the cosine similarity between the facial feature vectors and the standard feature templates; and determining a comprehensive judgment score based on the first similarity, the second similarity, and a preset maximum distance threshold.

[0010] In some embodiments, the central control platform is configured to respond to a device access request submitted by a user terminal, perform an access control review, and generate an access authorization command and a list of devices to be accessed when the access control review result is approved. This includes: parsing the device access request; the device access request includes: the applicant's identity identifier and the target device number; calling the administrator's permission database, and if the user is deemed qualified to apply based on the applicant's identity identifier, sending the application information to the administrator terminal for approval and confirmation; if the approval result returned by the administrator terminal is approved, querying the real-time status in the device inventory database, filtering out available devices to generate a list of devices to be accessed, and simultaneously generating an access authorization command containing the user's biometric index and the storage coordinates of the target device location.

[0011] In some embodiments, the warehouse management system further includes: an environmental monitoring and self-adjustment module, which includes a temperature and humidity sensor and an adjustment device; the adjustment device is a device for adjusting the temperature and humidity parameters inside the warehouse; the temperature and humidity sensor is used to collect the temperature and humidity signals of the warehouse at preset time intervals and convert them into current environmental values; the environmental monitoring and self-adjustment module is configured to: when the current environmental value is greater than or equal to a preset standard value, determine the operating power and working time required by the adjustment device based on the difference and rate of change between the current environmental value and the preset standard value, and issue adjustment instructions to the adjustment device.

[0012] In some embodiments, the central control platform is further configured to: collect multi-source sensor data in real time during the operation of the warehouse management system, the multi-source sensor data including at least access control switch status, warehouse temperature and humidity, equipment vibration signals, and network communication traffic; input the real-time collected multi-source sensor data into a preset anomaly detection model to determine a comprehensive anomaly score; when the comprehensive anomaly score exceeds a preset security baseline, determine that there is a potential intrusion or equipment failure, and perform a data circuit breaker operation; wherein, the data circuit breaker operation includes: cutting off external data connections, locking all equipment storage locations, and switching the warehouse management system to a local offline operation mode; the local offline operation mode is used to indicate that only the highest-privilege administrator is allowed to perform system recovery operations through a physical key or a local dedicated interface.

[0013] Secondly, this application also provides a warehouse management method including: responding to a device retrieval request submitted by a user terminal, performing permission review, and generating an access authorization instruction and a list of devices to be retrieved when the permission review result is passed; performing biometric verification based on the collected user facial image and access authorization instruction, and controlling the access control device to open the access control after the biometric verification result is passed; calculating the target path from the access control location to the target storage location according to the storage coordinates of the target storage location in the list of devices to be retrieved, and sequentially lighting up the display units of each device storage location on the target path according to the time delay to form dynamic light guidance; comparing the real-time tag code read by the reading device of the target storage location with the target device code in the list of devices to be retrieved, and controlling the electronic lock of the target storage location to perform a device release operation or trigger an abnormal prompt based on the comparison result.

[0014] Thirdly, this application also provides an electronic device comprising: a processor and a memory; the memory storing processor-executable instructions; when the processor is configured to execute the instructions, causing the electronic device to implement the method of the second aspect described above.

[0015] Fourthly, this application also provides a computer-readable storage medium comprising: computer software instructions; when the computer software instructions are executed in an electronic device, the electronic device performs the method described in the second aspect above.

[0016] Fifthly, this application also provides a computer program product comprising a computer program that, when run on an electronic device, causes the electronic device to perform the method of the second aspect.

[0017] The beneficial effects of the second to fifth aspects mentioned above can be referred to the corresponding description of the first aspect, and will not be repeated here. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A schematic diagram of the framework of a warehouse management system provided in this application; Figure 2 This application provides a schematic diagram of the access control process for a warehouse management system. Figure 3 This application provides a schematic diagram of a biometric verification process for a warehouse management system. Figure 4This application provides a schematic diagram of a warehouse management system for path calculation and dynamic lighting guidance. Figure 5 This application provides a schematic diagram of a warehouse management system for retrieval verification and hierarchical alarm processes; Figure 6 This application provides a schematic diagram of an inventory forecasting and replenishment early warning process for a warehouse management system. Figure 7 This application provides a schematic diagram of an anomaly detection and circuit breaker process for a warehouse management system. Figure 8 A schematic diagram of the framework of another warehouse management system provided in this application; Figure 9 A flowchart illustrating a warehouse management method provided in this application; Figure 10 This is a schematic diagram of the structure of an electronic device provided in this application. Detailed Implementation

[0020] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] It should be noted that in the embodiments of this application, the words "exemplarily" or "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design scheme described as "exemplarily" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of the words "exemplarily" or "for example" is intended to present the relevant concepts in a specific manner.

[0022] In the embodiments of this application, the terms "first," "second," "third," "fourth," "fifth," and "sixth" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," "third," "fourth," "fifth," and "sixth" may explicitly or implicitly include one or more of that feature.

[0023] In embodiments of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. For "A and / or B," this includes three combinations: A only, B only, and a combination of A and B.

[0024] With the increasing demand for refined security management of warehouses in logistics warehousing, power maintenance, and military field operations, traditional skid-mounted warehouses, while convenient for transfer and assembly and adaptable to various sites, suffer from inherent technical deficiencies in their intelligent management systems. For example, in terms of access control, existing access control systems generally use IC card recognition or mechanical key opening and closing schemes, resulting in limited identity verification methods, insufficient traceability of permissions, and inadequate anti-fraud capabilities, leading to a low overall security level. Regarding environmental management, existing environmental monitoring devices can only passively trigger alarms when fixed thresholds are exceeded, failing to implement closed-loop autonomous adjustments based on changes in parameters such as temperature, humidity, and gas levels within the warehouse, thus lacking dynamic adaptive capabilities. In terms of material and equipment management, the location, retrieval, and inventory of high-value equipment within the warehouse rely entirely on manual labor, resulting in cumbersome processes, low efficiency, and a high risk of human error leading to discrepancies between physical inventory and records. In summary, existing traditional skid-mounted warehouses cannot meet the modern, end-to-end management requirements for the secure storage, precise positioning, and efficient retrieval of high-value equipment.

[0025] Based on this, this application provides a warehouse management system, which includes: an access control device, multiple equipment storage locations, and a central control platform. Specifically, the central control platform first reviews user-submitted equipment retrieval requests for permission and generates access authorization instructions and a list of equipment to be retrieved, ensuring that only approved legitimate requests proceed to the next stage and preventing irrelevant requests from directly accessing the warehouse. Next, biometric verification is performed based on facial images captured by the access control device; the access control device is only opened after successful verification, ensuring that the actual person present matches the authorized identity and preventing identity theft. After entering the warehouse, the optimal path from the access control location to the target storage location is calculated based on the storage coordinates of the target storage location in the list of equipment to be retrieved, and each device along the path is illuminated sequentially according to time. The display unit for the storage location provides dynamic lighting guidance, allowing users to quickly and accurately reach the target location along the light strip without having to memorize or search, significantly improving retrieval efficiency and reducing the possibility of picking the wrong shelf. When a user retrieves equipment from the target storage location, the system reads the equipment's real-time tag code and compares it with the target equipment code in the inventory list. If they match, the electronic lock is activated to release the equipment; otherwise, an error message is triggered. This forms a closed-loop verification process, effectively preventing equipment from being taken by mistake, ensuring accurate inventory, and providing immediate alerts to personnel, thus enhancing the security and reliability of equipment storage and retrieval in the warehouse.

[0026] The warehouse management system provided in this application can be applied to precision parts warehouses, scientific research laboratories, military warehouses, hospital pharmacy equipment rooms, and special logistics warehouses; it can realize personnel access verification, automatic environmental control, rapid material location and inventory early warning, respectively ensuring the storage safety and management efficiency of precision components, experimental samples, military supplies, medicines and high-value goods.

[0027] Figure 1 This is a schematic diagram illustrating the framework of a warehouse management system provided in this application. Figure 1 As shown, the warehouse management system includes an access control device 101, multiple equipment storage locations 102, and a central control platform 103. The central control platform 103 is connected to the access control device 101 and the multiple equipment storage locations 102.

[0028] Access control device 101 is installed at the warehouse entrance and is used to collect user facial images.

[0029] The access control device 101 includes a face capture camera and an electronic lock. The face capture camera can be installed in front of the warehouse entrance to capture real-time facial images of people entering the warehouse. The electronic lock is linked to the warehouse door and unlocks the access control device after receiving an opening signal from the central control platform 103.

[0030] Multiple equipment storage positions 102 are respectively set inside the warehouse, and each equipment storage position 102 is used to store equipment to be retrieved.

[0031] Each device storage slot 102 includes a display unit, a reading device, and an electronic lock. The display unit is used to display guidance information; the reading device is used to read the tag code of the stored device; and the electronic lock is used to lock or release the stored device to be retrieved.

[0032] In one implementation, the display unit can be an LED indicator or a small display screen, installed on the front panel of a shelf or storage compartment, to display dynamic light guidance information based on the shortest path from the access control location to the target device storage location.

[0033] In one implementation, the reading device can be an RFID radio frequency identification reader, installed inside the storage compartment or at an opening, used to read the electronic tags attached to the stored device to obtain the tag code.

[0034] In one implementation, the electronic lock can be an electromagnetic lock or an electronically controlled latch, installed on the opening and closing part of the storage location, used to lock or release the stored device to be retrieved.

[0035] The central control platform 103 is an industrial control computer or server deployed locally or in the cloud, including a processor, memory and communication interface, responsible for data interaction and logical control with the access control device 101 and each device storage location 102.

[0036] In this embodiment, the central control platform 103 is configured to: respond to a user terminal's request for equipment retrieval, perform permission review, and generate an access authorization command and a list of devices to be retrieved upon successful review; perform biometric verification based on the user's facial image collected by the access control device 101 and the access authorization command, and control the access control device 101 to open the access control upon successful verification; calculate the target path from the access control location to the target storage location according to the storage coordinates of the target storage location in the list of devices to be retrieved, and sequentially light up the display units of each device storage location 102 on the target path according to the time delay to form dynamic light guidance; compare the real-time tag code read by the reading device of the target storage location with the target device code in the list of devices to be retrieved, and if they match, control the electronic lock to release the device, otherwise trigger an error message. Thus, through the coordinated operation of the access control device 101, each device storage location 102, and the central control platform 103, a fully intelligent closed-loop management system is achieved from permission review, biometric verification, dynamic path guidance to retrieval verification, effectively improving the efficiency and security of warehouse equipment retrieval.

[0037] The following describes specific embodiments of the warehouse management system of this application.

[0038] In some embodiments, please refer to Figure 2 , Figure 2This application provides a schematic diagram of a warehouse management system access control process. The central control platform is configured to respond to device retrieval requests submitted by user terminals, perform access control checks, and generate an access authorization command and a list of devices to be retrieved when the access control check is successful. This list includes: S201, Application for access to analytical equipment.

[0039] The equipment access application includes: the applicant's identity identifier and the target equipment number.

[0040] The applicant user identifier refers to information used to uniquely identify the applicant user, such as an employee ID, system registration account, or an index number bound to the user's biometric template. The target device number refers to a unique code for the device the user needs to retrieve. This code is assigned and bound to the device when the device is received and stored in the device inventory database.

[0041] In one implementation, the device retrieval application also includes: application time and the number of target devices requested. The application time is used to record the precise moment when the user submits the application, facilitating subsequent auditing and traceability; the number of devices requested indicates the number of devices the user needs to retrieve this time, so that the platform can filter out a sufficient number of available devices when generating the list of devices to be retrieved.

[0042] In some implementations, parsing device access requests refers to the central control platform disassembling the application data packets submitted by user terminals, extracting key fields such as user identification, target device number, application time, and application quantity, and converting them into structured data that can be called upon in subsequent review processes.

[0043] In some implementations, users can fill out and submit a device access request through an application installed on a mobile terminal or by scanning a QR code to access a webpage. The central control platform receives the request data packet through a wireless communication interface.

[0044] Therefore, by standardizing the parsing of user-submitted access applications, subsequent qualification reviews can quickly and accurately obtain key information, laying the foundation for automated approval processes.

[0045] S202. Call the administrator privilege database, and if the user is deemed qualified to apply based on the applicant's identity, send the application information to the administrator terminal for approval and confirmation.

[0046] The administrator privilege database refers to a collection of data that stores information such as the privilege levels of each user, the range of equipment categories that can be applied for, and historical violation records. This database is pre-configured by the system administrator and updated regularly.

[0047] In one implementation, the central control platform uses the extracted applicant user's identity identifier as a query condition to search for the user's corresponding permission level and eligible device category in the administrator permission database. If the target device number is within the user's permission range and the user has no outstanding violation records, the user is deemed qualified to apply. Otherwise, if the user's permissions are insufficient or there are violation restrictions, a rejection code is returned directly without proceeding to the administrator approval process.

[0048] Therefore, by automatically pre-screening applications before administrator approval, unqualified or risky applications can be effectively blocked, reducing the burden of manual approval while improving security.

[0049] S203. If the approval result returned by the administrator terminal is "passed", query the real-time status in the equipment inventory database, filter out available equipment to generate a list of equipment to be picked up, and at the same time generate a pass authorization instruction containing the user's biometric index and the storage coordinates of the target storage location.

[0050] An equipment inventory database refers to a database that stores information about all equipment in the warehouse. One implementation involves a database containing multiple records, each with fields such as equipment code, equipment name, specifications, storage location coordinates, and current status (available / occupied / under maintenance). This database is updated in real time when equipment is received or removed from the warehouse.

[0051] The list of devices to be retrieved refers to a list generated by the central control platform that contains information about the devices authorized for retrieval this time. The list includes at least the target device code, device name, and target storage location coordinates.

[0052] The access authorization command refers to a string of encrypted data packets generated by the central control platform, which may contain information such as the effective timestamp, the user's biometric index, and the storage coordinates of the target location. This command is used in the subsequent access control verification process to retrieve the corresponding preset biometric template for comparison.

[0053] In another possible implementation, if the administrator terminal returns an approval result of "not approved", a rejection code is generated and recorded in the audit log.

[0054] It should be understood that the list of devices to be retrieved and the access authorization instruction are only generated after the administrator actively approves them, ensuring that each retrieval operation is manually confirmed. At the same time, the user's biometric index is written into the authorization instruction, so that the corresponding template can be accurately retrieved in the subsequent verification process, avoiding the security risks of biometric templates being traversed and matched.

[0055] Based on S201-S203, the central control platform analyzes user-submitted equipment retrieval applications, extracting key fields such as the applicant's identity and the target equipment number. This enables subsequent qualification verification to quickly and accurately obtain the necessary information. The platform then automatically pre-screens user qualifications by accessing the administrator's permission database, blocking applications from unqualified applicants or those with violation records. This reduces the burden of manual approval while improving security. After the administrator terminal returns the approval result, the platform queries the equipment inventory database to filter available equipment and generate a list of equipment to be retrieved. Simultaneously, it generates an access authorization instruction containing the user's biometric index and the storage coordinates of the target equipment location. Thus, this application constructs a three-level closed-loop verification system: "application parsing—automatic pre-screening—manual approval—list and authorization instruction generation." This ensures that only legitimate applications that have undergone multiple verifications can proceed to subsequent access control verification and equipment retrieval stages, effectively balancing approval efficiency with the rigor of warehouse security control.

[0056] In some embodiments, please refer to Figure 3 , Figure 3 This application provides a schematic diagram of a biometric verification process for a warehouse management system. The central control platform is configured to perform biometric verification based on collected user facial images and access authorization commands. Upon successful biometric verification, the platform controls the access control device to open the door, including: S301. Extract the facial feature vector of the collected face image, compare it with the standard feature template pre-stored in the access authorization instruction, and determine the comprehensive judgment score.

[0057] A facial feature vector is a multidimensional numerical sequence extracted from a user's facial image to represent key facial features. Examples include geometric or deep learning features of key points such as eye spacing, nose bridge height, cheekbone contour, and lip shape.

[0058] In one implementation, the user's facial image is captured in real time by a face capture camera in the access control device. The central control platform performs preprocessing on the captured image, such as grayscale conversion, noise reduction and enhancement, and face region cropping. Then, a fixed-dimensional facial feature vector is extracted from the preprocessed image using a pre-trained convolutional neural network or local binary pattern feature extraction algorithm.

[0059] The standard feature templates pre-stored in the access authorization instructions refer to the facial biometric template data that users pre-enter during the system registration phase and that has been reviewed and confirmed by the administrator.

[0060] In one implementation, when a user registers for the first time, multi-angle facial images are collected through an access control device or a dedicated registration terminal. The system extracts feature vectors from the collected images and stores them in a biometric template database in an encrypted manner, generating a unique user biometric index. When S203 generates an access authorization command, the central control platform retrieves the corresponding standard feature template from the biometric template database based on the applicant user's identity and writes its index into the access authorization command for use in the access control verification process.

[0061] In some implementations, the central control platform is configured to extract facial feature vectors from the acquired face images, compare them with pre-stored standard feature templates in the access authorization instructions, and determine a comprehensive judgment score, including: determining a first similarity based on the Euclidean distance between the facial feature vectors and the standard feature templates; determining a second similarity based on the cosine similarity between the facial feature vectors and the standard feature templates; and determining a comprehensive judgment score based on the first similarity, the second similarity, and a preset maximum distance threshold.

[0062] One implementation method involves determining the first similarity based on the Euclidean distance between facial feature vectors and standard feature templates. This includes: calculating the square root of the sum of the squares of the differences between all facial feature vectors and their corresponding standard feature templates, which is then used as the Euclidean distance calculation to obtain the first similarity.

[0063] For example, the first similarity is determined to satisfy the following relationship:

[0064] in, This represents the i-th component of the facial feature vector extracted from the collected user face image. Let represent the i-th component of the standard feature template, where i ranges from 1 to n, and n represents the dimension of the feature vector. This formula represents the sum of the squares of the differences between the facial feature vector and the standard feature template in each dimension, and then taking the square root to calculate the Euclidean distance between them. This Euclidean distance is then used as the first similarity.

[0065] It should be understood that by directly quantifying the absolute distance between two feature vectors in space using Euclidean distance, the overall difference between the face image and the template can be intuitively reflected. The smaller the distance, the more similar the image is, providing a geometric distance-based metric for subsequent comprehensive judgment.

[0066] One implementation method involves determining a second similarity based on the cosine similarity between the facial feature vector and the standard feature template. This includes calculating the cosine value of the angle between the facial feature vector and the standard feature template. The closer the cosine value is to 1, the more consistent the directions of the two are. This cosine value is then determined as the second similarity.

[0067] For example, the second similarity is determined to satisfy the following relationship:

[0068] in, This represents the i-th component of the facial feature vector extracted from the collected user face image. Let represent the i-th component of the standard feature template, where i ranges from 1 to n, and n represents the dimension of the feature vector. The numerator is the sum of the products of the facial feature vector and each component of the standard feature template, and the denominator is the product of the magnitudes of the two vectors. This formula represents the cosine value of the angle between the two vectors, and this cosine similarity is determined as the second similarity.

[0069] It should be understood that using cosine similarity to measure the similarity of two vectors in a direction is not sensitive to changes in the overall brightness of the image. It can effectively compensate for the shortcomings of Euclidean distance under uneven lighting conditions and improve the stability of face comparison in different environments.

[0070] One implementation method involves determining a comprehensive judgment score based on a first similarity, a second similarity, and a preset maximum distance threshold, including: normalizing the first similarity and then weighting and summing it with the second similarity to obtain the comprehensive judgment score.

[0071] For example, the comprehensive judgment score is determined to satisfy the following relationship:

[0072] in, This is a preset maximum distance threshold used to normalize the Euclidean distance; and These are the weight coefficients for the first and second similarities, respectively, and they satisfy... ; Transform the Euclidean distance into a similarity metric that runs in the same direction as the cosine similarity.

[0073] It should be understood that by weighting and fusing the normalized distance and angle metrics, the comprehensive judgment score can simultaneously take into account the absolute positional and directional differences of the feature vectors, overcoming the limitations of a single similarity metric in scenarios with complex lighting and angular offsets, and significantly improving the accuracy of face verification.

[0074] Based on S301, by comparing the similarity of the facial feature vectors collected on-site with the standard feature templates pre-stored in the access authorization instructions and generating a comprehensive judgment score, the verification process is directly associated with the approved authorization instructions, avoiding the traversal and matching of the facial feature database. This improves the accuracy of verification while reducing the risk of false release.

[0075] S302. Perform liveness detection based on user action data and determine the liveness detection result.

[0076] User action data refers to real-time dynamic behavioral information generated by users during access control verification, including blinking sequences, head rotation angles, mouth opening and closing states, and facial micro-expression changes.

[0077] In one implementation, liveness detection is performed based on user action data to determine the liveness detection result. This includes: the access control device's face capture camera simultaneously captures multiple frames of the user's facial image sequence; the central control platform detects blinking by analyzing changes in eye opening and closing between adjacent frames; and a head pose estimation algorithm detects whether the user performs a nodding or shaking motion as required. If a motion sequence matching the instruction is detected within a preset time period, the liveness detection is deemed successful; otherwise, if the user remains stationary or their actions are inconsistent with the instruction, the liveness detection is deemed unsuccessful.

[0078] For example, after detecting a face, a set of unpredictable action command sequences is randomly generated and prompted to the user through the access control device's display screen or speaker, such as "Please blink first, then turn your head to the left, and finally open your mouth." The access control device's face capture camera continuously captures multiple frames of facial image sequences during the user's execution of the commands. The central control platform performs temporal analysis on each frame, sequentially detecting changes in the eye opening / closing ratio of blinking, changes in the yaw angle of head turning, and changes in the aspect ratio of mouth opening. If the user completes all actions in the specified order within a specified time, and the amplitude and duration of each action conform to the preset parameter range, then the liveness detection is deemed successful.

[0079] For example, the access control device captures the user's facial image while simultaneously capturing the user's voice signal via a microphone. The central control platform issues an instruction to the user, requiring them to read aloud a randomly generated string of numbers or a short sentence, while maintaining a smile or making a specified facial expression during the reading. The platform performs facial expression recognition on the captured facial image sequence, detecting whether micro-expression features such as the degree of mouth twitching and changes in eye wrinkles match the specified expression; simultaneously, it extracts voiceprint features from the voice signal and compares the extracted voiceprint features with the user's voiceprint template pre-stored in the access authorization instruction. If the facial expression recognition result matches the instruction requirements, and the voiceprint matching degree exceeds a preset threshold, then the liveness detection is deemed successful.

[0080] For example, the access control device is equipped with a near-infrared supplementary light and a near-infrared camera. During verification, the near-infrared supplementary light emits near-infrared light at a specific wavelength toward the user's face, and the near-infrared camera captures a near-infrared reflection image of the face. The central control platform extracts subcutaneous blood vessel distribution texture features or deep skin scattering texture features from the near-infrared image and compares these texture features with the user's near-infrared subcutaneous texture template pre-stored in the access authorization command. If the matching degree of the texture features exceeds a preset threshold, the liveness detection is deemed successful.

[0081] For example, the access control device's facial recognition camera continuously captures multiple frames of color images of the user's face at a preset frame rate. The central control platform performs pixel-level temporal analysis on regions of interest on the face (such as the forehead and cheeks), and extracts the invisible periodic color micro-change signals on the skin surface caused by heartbeats using Euler video magnification algorithms or remote photoplethysmography technology, and calculates the pulse wave waveform and heart rate value accordingly. If a periodic pulse wave signal that conforms to the normal human heart rate range is extracted, the liveness detection is deemed successful.

[0082] Based on S302, by introducing dynamic motion detection such as blinking and head turning during the verification process, the system can effectively distinguish between real human bodies and forgery attacks such as photos, videos or masks, making up for the shortcomings of static face comparison in anti-spoofing capabilities and further strengthening the access control security defense.

[0083] S303. If the comprehensive judgment score is greater than the preset threshold and the liveness detection result meets the preset conditions, determine that the biometric verification result is passed and control the access control device to open the access control.

[0084] The preset conditions include: the liveness detection determines that the user is a real human body, that is, the user has completed the specified action sequence according to the random instructions of the system and the timing analysis results of each action parameter are within the preset range.

[0085] One approach is to set a preset threshold based on the system’s actual security requirements. The higher the security requirements, the closer the preset threshold is to the maximum score. Conversely, in scenarios with low security requirements, the threshold can be appropriately reduced to improve passage efficiency.

[0086] For example, if the comprehensive judgment score is greater than a preset threshold and the liveness detection result meets preset conditions, the biometric verification result is determined to be passed, and the access control device is controlled to open the access control, including: the central control platform obtains the comprehensive judgment score calculated by S301. and the liveness detection results determined by S302; With preset threshold The comparison is performed, and it is determined whether the liveness detection result is satisfactory; if... If the liveness detection result is successful, a biometric verification pass signal is generated, and an opening command is sent to the electronic lock of the access control device to unlock the warehouse access control. Simultaneously, the verification pass timestamp and user identity information are recorded in the audit log. If any condition is not met, a verification failure signal is generated, access control is refused, and an abnormal alarm message is pushed to the administrator terminal. The warehouse can be dynamically adjusted according to the value level of the equipment stored there; high-value equipment can be stored in a separate warehouse. The value should be set to 0.9 or higher. For ordinary material warehouses, the value can be set between 0.7 and 0.8.

[0087] Based on S303, by binding the comprehensive judgment score threshold with the liveness detection result as a dual condition, the access control is only opened under the dual premises of the face similarity meeting the standard and the person being confirmed as a live person. This avoids security risks caused by a single condition being met and significantly enhances the reliability of access control authorization.

[0088] Based on S301-S303, the central control platform first extracts facial feature vectors from the user's face image and performs multi-dimensional similarity comparison with the standard template pre-stored in the authorization command to generate a comprehensive judgment score. Then, it combines dynamic action data driven by random commands to perform liveness detection. Finally, the access control is opened under the dual conditions of a satisfactory comprehensive score and successful liveness detection. Thus, this application constructs a composite biometric verification mechanism of "static feature comparison + dynamic liveness detection," which avoids full traversal matching of the facial feature database and effectively resists forgery attacks such as photos and videos through random command sequences, achieving a dual improvement in the accuracy and security of identity verification.

[0089] In some embodiments, please refer to Figure 4 , Figure 4 This application provides a schematic diagram of a path calculation and dynamic lighting guidance process for a warehouse management system. The central control platform is configured to calculate the target path from the access control location to the target storage location based on the storage coordinates of the target storage location in the list of devices to be retrieved, specifically including: S401. Based on the two-dimensional electronic map of the warehouse, the shortest walking path is determined by taking the access control location as the starting node and the target storage location as the ending node.

[0090] The warehouse's two-dimensional electronic map is a floor plan of the warehouse's interior, pre-stored in the central control platform. The floor plan is marked with the physical coordinates of each equipment storage location and the passage distance between each adjacent node.

[0091] In one implementation, a weighted undirected graph is constructed as a two-dimensional electronic map of the warehouse, with each equipment storage location and passageway intersection as nodes, the passable passages between adjacent nodes as edges, and the actual walking distance between nodes as the edge weights.

[0092] A preset path algorithm is used to search a 2D electronic map from a starting node, expanding the search level by level to the ending node, and selecting the path with the shortest total distance. In one implementation, the preset path algorithm uses Dijkstra's shortest path algorithm, parses the list of devices to be retrieved, obtains the physical storage coordinates of the target devices and the user's current access control coordinates, and calculates the optimal path cost. The following relationship must be satisfied:

[0093] in, This represents the number of hops in the path. This represents the distance between adjacent nodes. The central control platform uses the access control coordinates as the starting node and the target storage coordinates as the ending node. The distance of the starting node is initialized to 0, and the distances of the other nodes are infinite. The platform selects the node with the smallest distance from the unvisited nodes and expands level by level, updating the shortest distance between adjacent nodes, until the ending node is visited. The platform then backtracks to obtain the node sequence with the smallest total travel distance.

[0094] Based on S401, by modeling the warehouse layout as a weighted undirected graph and using the shortest path algorithm to automatically calculate the optimal passage path, the system can plan the shortest walking route for users each time they need to retrieve equipment without relying on experience or manual searching. This reduces the time users spend walking around the warehouse and improves retrieval efficiency.

[0095] S402. Decompose the target path into a sequence of nodes, and sequentially light up the display units of the corresponding device storage positions of each node from the starting node to the ending node according to the preset time interval.

[0096] In one implementation, the optimal path is decomposed into Path node sequence It generates sequential lighting control commands, corresponding one-to-one with the display screens in each storage location, and defines the timing delay for the light guidance. Calculate the first When the guide light illuminates The following relationship must be satisfied:

[0097] in, This marks the initial lighting time. The central control platform controls the display units in each storage location to... The lights illuminate sequentially, creating a dynamic light guide that directs the user along the optimal path to the target device's storage location. After the user arrives and retrieves the device, the system receives a status change signal from the device's on-site sensor, controls the corresponding display screen to turn off, and clears the queue of devices awaiting guidance for that path.

[0098] In one implementation, when the list of devices to be retrieved contains multiple target storage locations, the central control platform sorts the multiple target storage locations according to the principle of proximity or concentration from the same channel direction, and connects the sorted target storage locations to the same target path to generate a one-time light guidance sequence. Users can follow the light guidance to retrieve all devices at once without having to repeatedly return to the access control to replan the path.

[0099] Based on S402, by converting the optimal path into a sequential lighting sequence with a time delay, users can follow the light guidance to naturally walk to the target storage location after passing through the access control, without having to stop in the warehouse to identify the direction or look up the shelf number, which significantly improves the accuracy of retrieval and the convenience of operation.

[0100] Based on S401-S402, the central control platform first models the physical layout of the warehouse as a weighted two-dimensional electronic map. It then uses a shortest path algorithm to automatically calculate the optimal path from the access control to the target storage location. The path is then decomposed into a sequence of nodes, and the display units along the path are illuminated sequentially according to a time delay, forming a dynamic, flowing light guide. Thus, this application achieves a technological leap from "static location query" to "dynamic path navigation." In emergency retrieval scenarios, users can quickly and accurately reach the target location along the light strip, significantly shortening the equipment retrieval response time. Simultaneously, the intuitiveness of the light guide effectively reduces mis-retrieval and delays caused by taking the wrong path.

[0101] In some embodiments, please refer to Figure 5 , Figure 5 This application provides a schematic diagram of a warehouse management system for retrieval verification and tiered alarm processes. The central control platform is configured to compare the real-time tag code read by the reading device at the target storage location with the target device code in the list of devices to be retrieved. Based on the comparison result, the platform controls the target storage location to perform a device release operation or trigger an anomaly alert, including: S501. If the real-time tag code and the target device code are consistent, control the target storage location to open and release the device, and record the successful device retrieval event.

[0102] In one implementation, the central control platform detects a user's arrival at the target storage location using an infrared sensor or pressure sensor located there, and then triggers an RFID reader or barcode scanner at that location to collect the tag code of the device the user has retrieved from the storage location. .Will Matching the target device code in the list of devices to be retrieved Perform the comparison, calculate the matching degree, and define it as a binary decision function:

[0103] when At that time, the central control platform controls the smart lock of the storage location to open, allowing the user to take the device. At the same time, a "successful retrieval" event data is generated and uploaded to the central control platform, and the retrieval time, device ID and operation user information are recorded in the audit log.

[0104] Based on S501, by accurately comparing the device tag codes collected on-site with the preset target device codes in the list of devices to be retrieved, the lock is controlled to release the device only when the two are completely consistent, ensuring a strict correspondence between the retrieval operation and the authorized list, and preventing the wrong retrieval of devices from the source.

[0105] S502. In the event of a discrepancy between the real-time tag encoding and the target device encoding, determine the alarm level based on the feature similarity between the device in the target storage location and the target device corresponding to the target device encoding; based on the alarm level, control the alarm volume and the alarm information push range.

[0106] In one implementation, when At that time, that is The central control platform determines that the currently retrieved device is inconsistent with the target device and calculates the feature similarity between the mistakenly retrieved device and the target device. This will trigger a tiered alarm. Alarm Level The following relationship must be satisfied:

[0107] in, , The closer the value is to 1, the more similar the characteristics of the mistakenly taken device are to the target device. The value is a positive integer rounded up, which determines the alarm volume level and message push level. The alarm information is simultaneously sent to the administrator terminal and the on-site audible and visual alarms, and a graphic instruction stating "Device incompatible, please return to its original location" is displayed on the screen at the current storage location.

[0108] For example, if the target device is a precision bearing type A, and the user mistakenly selects a precision bearing type B that has a highly similar appearance, the feature similarity between the two is... The alarm level is... At level 1, trigger the highest volume alarm and push alerts to administrators at all levels; if the user mistakenly selects a completely unrelated general tool, ,but Level 1: Only triggers a low volume alert and pushes a reminder to the administrator on duty.

[0109] One implementation involves calculating the feature similarity between the mistakenly taken device and the target device. This includes: a central control platform that pre-maintains a database of equipment feature attributes, storing multi-dimensional feature attributes such as function classification, specification level, and physical dimensions for each piece of equipment. When a mismatch is determined, the database is used to retrieve the feature attributes of both the target equipment and the incorrectly taken equipment. The similarity between the two in terms of function classification, specification level, and size parameters is compared across dimensions. A weighted sum of the individual similarities for each dimension is then used to obtain a comprehensive feature similarity value ranging from 0 to 1. The closer this value is to 1, the more similar the devices are.

[0110] Based on S502, by dynamically linking the feature similarity of mistakenly taken devices with the alarm level, the system can automatically adjust the alarm intensity according to the severity of the mistake. For high-risk mistakenly taken cases with similar appearances but different specifications, a higher level alarm is triggered, avoiding the problem of insufficient warning for high-risk mistakenly taken events by traditional single alarm methods.

[0111] Based on S501-S502, this application constructs a closed-loop verification mechanism for tag acquisition, consisting of "real-time tag acquisition—binarization matching judgment—tiered alarm response." When the tags match, access is automatically granted and recorded; when the tags do not match, the alarm level is dynamically calculated based on the similarity of features between devices. This achieves a technical upgrade from "one-size-fits-all alarms" to "risk-tiered response," ensuring accuracy while also considering operational efficiency.

[0112] In some embodiments, the warehouse management system further includes an environmental monitoring and self-regulation module.

[0113] The environmental monitoring and self-regulation module includes temperature and humidity sensors and regulation devices.

[0114] In some implementations, the regulating device is a device that regulates the temperature and humidity parameters inside the warehouse, such as an air conditioner, dehumidifier, or humidifier; the temperature and humidity sensor is used to collect the temperature and humidity signals of the warehouse at preset time intervals and convert them into current environmental values.

[0115] It should be understood that by deploying temperature and humidity sensors in the warehouse to collect environmental data in real time, continuous and accurate monitoring data is provided for subsequent automatic adjustments, replacing the lag in manual inspections.

[0116] In some implementations, the environmental monitoring and self-adjustment module is configured to: when the current environmental value is greater than or equal to a preset standard value, determine the operating power and working time required by the adjustment device based on the difference and rate of change between the current environmental value and the preset standard value, and issue adjustment instructions to the adjustment device.

[0117] In one implementation, the environmental monitoring and self-adjustment module uses a PID control algorithm to calculate the adjustment amount. The deviation between the current environmental value and the preset standard value is used as the proportional term input, the cumulative amount of the deviation over time is used as the integral term input, and the rate of change of the deviation is used as the derivative term input. The power control amount of the regulating device is obtained by weighted summing of the three terms, so as to achieve precise and smooth adjustment of temperature and humidity and avoid over-adjustment or oscillation.

[0118] In one implementation, after the environmental monitoring and self-adjustment module sends an adjustment command to the adjustment device, it collects environmental data again within a preset delay time and calculates the deviation reduction rate. If the deviation reduction rate does not reach the expected target, it automatically increases the power level of the adjustment device or extends the working time until the environmental parameters are restored to the preset standard value range.

[0119] In one implementation, independent temperature and humidity sensors are deployed in different areas of the warehouse. The environmental monitoring and self-adjustment module generates differentiated adjustment instructions for different areas based on the environmental values ​​returned by the sensors in each area, so as to make the environmental parameters in each area of ​​the warehouse balanced and stable, and avoid local temperature and humidity differences caused by different equipment storage densities.

[0120] It should be understood that by combining PID closed-loop automatic adjustment with zoned differentiated control, the temperature and humidity in the warehouse are kept stable within the preset range. This avoids the problem of delayed manual intervention when the environment deviates from the standard, which is a problem in the traditional single threshold alarm mode. This effectively ensures the storage quality of environmentally sensitive equipment such as precision instruments and experimental samples.

[0121] In some embodiments, please refer to Figure 6 , Figure 6 This application provides a schematic diagram of an inventory forecasting and replenishment early warning process for a warehouse management system. The central control platform is also configured as follows: S601. Construct a time series of equipment consumption based on historical inventory data.

[0122] Historical inventory data refers to the data on equipment entry, exit, and inventory balance changes recorded by the warehouse management system during its past operation. Each record includes a timestamp and the corresponding inventory value at that time.

[0123] In one implementation, the central control platform aggregates and statistically analyzes historical outbound records according to a preset time granularity (such as hourly, daily, or weekly), calculates equipment consumption within each time interval, and arranges them in chronological order to construct an equipment consumption time series. ,in Let be the inventory or consumption amount for the t-th time interval. Before constructing the time series, the raw data is also imputed for missing values ​​and smoothed for outliers to eliminate noise interference caused by sensor malfunctions or data entry errors, ensuring that the series data can accurately reflect the consumption patterns of the equipment.

[0124] Based on S601, by constructing the scattered historical outbound records into a standardized time series, the discrete inventory change information is transformed into a time series data structure that can be learned by the neural network model, laying the data foundation for subsequent predictive modeling.

[0125] S602. Input the equipment consumption time series into the preset neural network model to perform multi-step rolling prediction and output the predicted inventory value at multiple future times.

[0126] In one implementation, the pre-defined neural network model can be a long short-term memory (LSTM) neural network model. The constructed device consumption time series is input into the LSTM model. The model uses gating mechanisms such as internal forget gates, input gates, and output gates to extract features and model the long-term dependencies in the time series. The computational logic of the LSTM unit is as follows: First, the LSTM unit receives the input at the current time step at each time step. (i.e., the inventory level in the t-th time interval of this application) and the hidden state at the previous time step. (That is, the feature encoding of the inventory change pattern in the previous time step is a vector with a preset dimension, the size of which is predetermined by the model structure), through the forget gate. (The output of the forget gate is a vector ranging from 0 to 1, where each component represents the proportion of historical information retained in the corresponding dimension.) This determines the cell state at the previous time step. (The cell state from the previous time step, used for long-term memory to store historical time-series information, is a vector with the same dimension as the hidden state.) Which historical inventory information needs to be discarded? The forget gate calculation formula is:

[0127] in, The weight matrix of the forget gate (is the parameter matrix that needs to be learned during model training, and its dimension is determined by the input dimension and the hidden state dimension). This is the bias vector for the forget gate (a parameter vector that needs to be learned during model training, with the same dimension as the hidden state). This means concatenating the hidden state vector from the previous time step with the current input (i.e., linking the two vectors end to end to form a new long vector). The sigmoid activation function outputs values ​​between 0 and 1, where 0 represents complete forgetting and 1 represents complete retention. In this application, when there is a one-time abnormal fluctuation in historical inventory data (such as a large-scale temporary transfer out of the warehouse), the forget gate can learn to gradually attenuate its impact, avoiding the continuous interference of outliers with subsequent predictions.

[0128] Secondly, the input gate (The output value of the input gate is a vector ranging from 0 to 1, where each component represents the proportion of new candidate information written in the corresponding dimension.) This determines the input at the current time step. How much new information needs to be written into the cell state? Simultaneously, candidate cell states are generated through the tanh layer. This refers to the newly generated candidate information at the current moment, with values ​​ranging from -1 to 1, representing new inventory change features extracted from the current input. The calculation formula is:

[0129]

[0130] in, The weight matrices for the input gate and candidate cell states are separate (both are parameter matrices that the model needs to learn). Here, is the bias vector, and tanh is the hyperbolic tangent activation function with an output value between -1 and 1. In this application, when the warehouse shows a trend of accelerated equipment consumption, the input gate will increase its opening degree to incorporate this new trend into the cell state.

[0131] Then, cell state (The cell state at the current moment, updated by combining historical long-term memory and newly written candidate information) is updated according to the control of the forget gate and the input gate, achieving selective retention of historical memory and fusion of new information:

[0132] in, This represents the Hadamard product (i.e., element-wise multiplication, where corresponding elements of two vectors of the same dimension are multiplied). In this application, This is equivalent to the model having a comprehensive long-term memory of the equipment inventory consumption patterns up to the current moment, reasonably forgetting old information and selectively absorbing new information, enabling the model to adaptively track changes in consumption trends.

[0133] Next, output gate (The output value of the output gate is a vector ranging from 0 to 1, where each component represents the proportion of cell state information in the corresponding dimension that is output to the hidden state.) Based on the current input Hidden state from the previous moment The decision is based on the current cell state. Which information is extracted as the output at the current moment? In this application, the output gate determines which parts of the current inventory change characteristics are useful for predicting future inventory levels. The output gate calculation formula is:

[0134] in, This is the weight matrix of the output gate. This is the bias vector for the output gate. This is the Sigmoid activation function, with output values ​​between 0 and 1.

[0135] Finally, the control result of the output gate is combined with the cell state activated by tanh to generate the hidden state at the current moment. In this application, This is the feature encoding of the model output at the current moment, which is the comprehensive feature vector of inventory changes extracted at this moment. It will then be mapped to the predicted inventory value through a fully connected layer. This vector contains fused information of filtered short-term fluctuation features and long-term trend features. The formula for calculating hidden state is:

[0136] During model training, the weight matrices in the LSTM model are optimized by minimizing the loss function. and bias vector The loss function satisfies the following relationship:

[0137] in, This represents the actual inventory value of the i-th sample (i.e., the actual inventory quantity recorded at a certain point in history in this application). This represents the model's predicted inventory value for that moment. This represents the total number of training samples (i.e., the number of historical data samples used in model training). , which is the regularization coefficient (i.e., a pre-defined hyperparameter used to balance the relationship between prediction error and model complexity). W represents the sum of the squares of the L2 norms of all weight matrices (i.e., the sum of the squares of each element in all weight matrices, used to constrain the size of the weights to prevent overfitting, where W refers to all weight matrices in the model). During prediction, a sliding window approach is used for multi-step rolling prediction, that is, the predicted value at the current time step is used as the input for the next time step, and the predicted inventory value at the next T time steps is obtained through progressive extrapolation. .

[0138] It should be understood that all formulas in this implementation correspond to the complete forward propagation and training logic of the LSTM model in the warehouse inventory prediction scenario. Among them, the forget gate... and input gate The balance between the forgetting of historical inventory information and the writing of new consumption features is controlled collaboratively, and the cell state is maintained. As a long-term memory unit, the output gate continuously tracks the consumption patterns of the device. and hidden state The long-term memory is transformed into the current moment's inventory change features and mapped to specific predicted values; the loss function... The model complexity is constrained by minimizing the mean squared error between the predicted and actual values ​​and introducing a regularization term. Thus, the formulas form a complete closed loop from information filtering, state updating, feature output to parameter optimization, enabling the model to automatically learn the periodic patterns and sudden trends in inventory consumption data, providing an accurate time-series modeling foundation for multi-step rolling forecasting.

[0139] Based on S602, the retention and updating of historical information are controlled by the forget gate and input gate of the LSTM model. This can effectively capture the long-term periodic patterns and sudden change trends in equipment consumption data. Compared with traditional moving average or linear regression methods, it significantly improves the prediction accuracy. The multi-step rolling prediction method enables the system to perceive the inventory status at multiple future moments in advance.

[0140] S603. When at least two of the predicted inventory values ​​at multiple future time points are lower than the corresponding preset safety stock threshold, a replenishment warning signal is generated and pushed to the administrator terminal.

[0141] In some implementations, the predicted inventory value at multiple future time points refers to the sequence of predicted future inventory values ​​output by the LSTM model in S602 through multi-step rolling prediction. The sequence is arranged sequentially according to a preset time step, and each value corresponds to the inventory forecast at a future time point.

[0142] For example, if the preset time granularity is daily, then This indicates the projected inventory level at the end of tomorrow. This represents the predicted inventory level at the end of the day after tomorrow, and so on. If at least two predicted values ​​in the sequence are lower than the corresponding safety stock threshold, it indicates that the inventory decline trend is continuous, and the system generates a replenishment warning signal accordingly.

[0143] Based on S601-S603, this application transforms inventory management from passive post-event replenishment to proactive pre-event prevention. It improves prediction accuracy through the gating mechanism of the LSTM model and effectively filters false alarms caused by single-point fluctuations through the multi-time joint judgment mechanism. This enables managers to be aware of stockout risks in advance and take countermeasures, effectively avoiding operational interruptions caused by inventory shortages and improving the continuity and reliability of warehouse material support.

[0144] In some embodiments, please refer to Figure 7 , Figure 7 This application provides a schematic diagram of an anomaly detection and circuit breaker process for a warehouse management system. The central control platform is also configured as follows: S701: Real-time acquisition of multi-source sensor data during the operation of the warehouse management system.

[0145] Among them, multi-source sensor data includes at least access control switch status, warehouse temperature and humidity, equipment vibration signals, and network communication traffic.

[0146] In some implementations, the access control switch status is used to reflect whether the warehouse access control has been abnormally opened or has not been closed for a long time.

[0147] In some implementations, warehouse temperature and humidity data are provided by an environmental monitoring module to monitor for signs of environmental out-of-control conditions.

[0148] In some implementations, equipment vibration signals are collected by vibration sensors deployed at each equipment storage location to detect whether there is forced demolition or abnormal equipment movement.

[0149] In some implementations, network communication traffic is used to reflect whether the system is under external threats such as network scanning or denial-of-service attacks.

[0150] In some implementations, the central control platform uses a pre-defined unified data acquisition interface to periodically poll or collect data from various sensors through events, aggregating multi-source heterogeneous data into a unified format multi-dimensional feature vector for subsequent anomaly detection models to use.

[0151] Based on S701, by integrating multi-dimensional sensor data such as access control, environment, vibration and network, the system can comprehensively perceive the warehouse operation status from multiple dimensions such as physical intrusion, environmental anomalies, equipment damage and network attacks, making up for the shortcomings of single data source in anomaly detection, such as limited coverage and easy to miss.

[0152] S702. Input the real-time collected multi-source sensor data into the preset anomaly detection model to determine the comprehensive anomaly score.

[0153] In one implementation, the preset anomaly detection model is the Isolation Forest algorithm. This algorithm, based on the distribution characteristics of multi-source sensor data under normal conditions, constructs multiple random binary trees during the training phase, recursively and randomly partitioning the feature space to isolate normal data points. During real-time detection, newly acquired multi-source sensor data points are traversed through each isolated tree, and the path length of each data point in each tree is calculated. The average path length is normalized and converted into a comprehensive anomaly score. The greater the deviation of the current feature combination of the multi-source sensor data from the normal state, the easier it is for the data point to be isolated; the shorter the path length, the higher the comprehensive anomaly score.

[0154] For example, when an access control system is opened during an unauthorized period, the vibration sensor in the corresponding storage location detects continuous abnormal vibration, and network traffic experiences abnormal peaks, the combination of features from the three sensor data points deviates significantly from the normal state. The Isolation Forest model will then assign a high overall anomaly score to this data point. Conversely, if only temperature and humidity data show brief, slight fluctuations while all other data are normal, the overall characteristics differ less from the normal state, resulting in a lower overall anomaly score and preventing misjudgment as an intrusion event.

[0155] Based on the S702, the system uses an isolated forest model to fuse and score multidimensional sensor data, enabling it to comprehensively judge abnormal signals from different sensor channels. This avoids false alarms caused by noise from a single sensor or instantaneous fluctuations, thus improving the accuracy and robustness of anomaly detection.

[0156] S703. When the overall anomaly score exceeds the preset security baseline, a potential intrusion or equipment malfunction is identified, and a data circuit breaker operation is performed.

[0157] The data circuit breaker operation includes: cutting off external data connections, locking all device storage locations, and switching the warehouse management system to local offline operation mode. The local offline operation mode is used to indicate that only the highest-privilege administrator is allowed to perform system recovery operations through a physical key or a local dedicated interface.

[0158] In some implementations, data circuit breaking operations include: disconnecting external data connections to isolate the system from external networks and prevent attackers from remotely controlling the system or leaking data; locking electronic locks on all device storage locations to prevent unauthorized removal of devices; and switching the warehouse management system to a local offline operating mode. The local offline operating mode instructs only the highest-privilege administrator to perform data export and system recovery operations via physical keys or a dedicated local interface, ensuring that troubleshooting and system restarts are completed in a secure environment away from the network.

[0159] In some implementations, a preset safety baseline is set based on warehouse safety level requirements and historical data statistical characteristics.

[0160] Based on S703, by comparing the comprehensive anomaly score with the security baseline threshold, three synchronous circuit breaker measures are immediately implemented when a major anomaly is confirmed: external network disconnection, full database locking, and offline switching. This enables the system to quickly enter a secure isolation state in the early stages of an attack or major failure, maximizing the containment of threat spread and providing a controlled local recovery environment for administrators to handle subsequent actions.

[0161] Based on S701-S703, the central control platform constructs a panoramic perception capability of the system's operational status by fusing multi-source sensor data. It utilizes the Isolation Forest algorithm to perform comprehensive anomaly scoring on multi-dimensional data, and automatically executes tiered circuit breaker operations from the network and devices to the system when the score exceeds the security baseline. Thus, this application constructs a proactive defense closed-loop mechanism of "multi-dimensional perception—fusion scoring—tiered circuit breaker," providing the warehouse management system with end-to-end security protection capabilities from threat detection to emergency blocking, effectively preventing security risks caused by network attacks, physical intrusions, and equipment failures.

[0162] In some embodiments, the warehouse management system provided in this application further includes a data security processing module, which is configured to provide end-to-end security protection for transmitted data. Specifically, this includes: when generating access authorization instructions and a list of devices to be retrieved, using the AES-256 encryption algorithm to encrypt sensitive fields such as user identification, biometric index, and target storage location coordinates to ensure the confidentiality of core authorization data during storage and transmission; simultaneously, establishing a blockchain node to package the user's retrieval application time, administrator review record, user biometric hash value, and entry / exit timestamps into blocks, and broadcasting them to the consortium blockchain for evidence storage, so that the entire process log of each retrieval operation has the characteristics of being tamper-proof and traceable; in addition, during the transmission of access control opening signals and device guidance signals, using the SSL / TLS protocol to establish an encrypted communication channel, and performing CRC cyclic redundancy check on each data packet. If the check fails, the data packet is actively discarded and a retransmission is requested from the sender, thereby effectively preventing data from being tampered with or intercepted during transmission. Therefore, this application constructs a data security protection system covering the entire process of data generation, storage, transmission and notarization by deploying three mechanisms in a coordinated manner: AES-256 encryption, blockchain notarization, and SSL / TLS secure transmission. This provides a solid guarantee for the safe and reliable operation of the warehouse management system.

[0163] In summary, the warehouse management system provided in this application achieves intelligent closed-loop management of the entire warehouse process through hardware collaboration of access control devices, multiple equipment storage locations, and a central control platform, along with functional modules such as access control, biometric verification, dynamic lighting guidance, closed-loop verification of retrieval, automatic environmental adjustment, intelligent inventory prediction, and proactive anomaly defense.

[0164] This application also provides another warehouse management system; please refer to [link / reference]. Figure 8 The other warehouse management system includes: a user interaction and authorization management module 801, a biometric verification module 802, an environmental monitoring and self-regulation module 803, an intelligent positioning and guidance module 804, a replenishment module 805, and a control platform 806.

[0165] It should be understood that, Figure 8 The system shown is Figure 1 The systems shown represent two different architectural expressions under the same inventive concept, and they correspond to each other in function: Figure 1 Based on the configuration logic of physical hardware such as access control device 101, equipment storage space 102 and central control platform 103, this paper focuses on describing the connection relationship and control process of hardware entities. Figure 8 It is presented in the form of software functional modules, and each functional module can be deployed in the control platform 806 for collaborative use. Figure 1 The system incorporates hardware resources such as access control devices 101 and equipment storage spaces 102 to achieve the same intelligent warehouse management functions. This is in line with... Figure 1 The central control platform 103 is distinguished from the central control platform. Figure 8 The name 806 is used to emphasize its role as the deployment carrier and scheduling center for various functional modules in the modular architecture.

[0166] Specifically, the user interaction and authorization management module 801 corresponds to Figure 1 The central control platform 103's permission review and authorization instruction generation function corresponds to the biometric verification module 802. Figure 1 The facial recognition logic of the central access control device 101 and the biometric verification logic of the central control platform 103 correspond to the environmental monitoring and self-regulation module 803. Figure 1 The temperature and humidity sensors and control equipment deployed in the central warehouse, along with their closed-loop control functions, and the corresponding intelligent positioning and guidance module 804, are... Figure 1 The collaborative control logic of the display unit, reading device, and electronic lock in each device storage position 102, and the corresponding replenishment module 805. Figure 1 The central control platform 103 features inventory forecasting and replenishment early warning functions. Therefore, Figure 8 and Figure 1 The technical solution of this application is described from both the perspective of software function and the perspective of hardware entity. The two can be referred to and understood together, and can be flexibly deployed in actual implementation.

[0167] The user interaction and authorization management module 801 is configured to receive device access requests submitted by user terminals, parse the user identity identifier and target device number in the request, call the administrator permission database for qualification pre-screening, and generate access authorization instructions and a list of devices to be accessed after the administrator approves the request.

[0168] The biometric verification module 802 is configured to collect user face images, extract facial feature vectors and perform dual comparisons of Euclidean distance and cosine similarity with the standard feature templates pre-stored in the access authorization command to generate a comprehensive judgment score. At the same time, it combines action data driven by random commands to perform liveness detection. After the comprehensive score meets the standard and the liveness detection is passed, it controls the access control to open.

[0169] The environmental monitoring and self-regulation module 803 includes a temperature and humidity sensor and a regulating device. It is configured to collect warehouse environmental parameters in real time, use a PID control algorithm to calculate the regulation amount and send it to the regulating device to keep the temperature and humidity stable within a preset range.

[0170] The intelligent positioning and guidance module 804 is configured to calculate the optimal retrieval path based on the coordinates of the target storage location in the list of devices to be retrieved, using the shortest path algorithm on the two-dimensional electronic map of the warehouse, and to sequentially light up the display units of each storage location on the path according to a preset time delay, forming dynamic light guidance; when the user retrieves the device, the module reads the device to obtain the real-time tag code and compares it with the target device code. If the comparison matches, the module controls the electronic lock to release the device; if they do not match, the module calculates the feature similarity between the mistakenly retrieved device and the target device and dynamically generates a graded alarm command.

[0171] The replenishment module 805 is configured to construct a time series of equipment consumption based on historical inventory data, and use an LSTM long short-term memory neural network model to perform multi-step rolling prediction. When the predicted inventory value at multiple future moments is continuously lower than the safety threshold, a replenishment warning is automatically generated.

[0172] The control platform 806 is connected to the signals of each of the above modules and is used to coordinate data interaction and logic control.

[0173] Thus, the system integrates functions such as access control, identity verification, environment adjustment, path guidance, retrieval verification, and inventory forecasting into a unified intelligent management platform through a modular architecture. The modules work together to achieve intelligent closed-loop management of the entire warehouse process.

[0174] The following is a specific embodiment to illustrate the detailed logic implementation of the warehouse management method of this application.

[0175] Figure 9 As shown, Figure 9 This application provides a flowchart illustrating a warehouse management method, which includes: S901. In response to the device access request submitted by the user terminal, perform permission review, and generate an access authorization instruction and a list of devices to be accessed when the permission review result is passed.

[0176] It should be understood that this step involves parsing the user's submitted request, extracting key fields such as the applicant's identity and the target device number, automatically pre-screening the request by calling the administrator's permission database, and then querying the device inventory database to filter available devices after administrator approval. Finally, it generates a list of devices to be retrieved and an authorization instruction containing the user's biometric index, thus constructing a multi-level closed-loop review process: "application parsing—automatic pre-screening—manual approval—authorization generation." For a detailed description of permission review in the above system embodiment, please refer to that description; it will not be repeated here.

[0177] S902. Based on the collected user face image and access authorization command, perform biometric verification, and control the access control device to open the access control after the biometric verification result is passed.

[0178] It should be understood that this step extracts facial feature vectors from the user's face image and performs a dual comparison of Euclidean distance and cosine similarity with the standard feature templates pre-stored in the access authorization command to generate a comprehensive judgment score. Simultaneously, it combines random command-driven blinking, head turning, and other action data for liveness detection. Access control is activated under the dual conditions of a satisfactory comprehensive score and successful liveness detection, thus constructing a composite biometric verification mechanism of "static feature comparison + dynamic liveness detection." For a detailed description of biometric verification in the above system embodiments, please refer to them; they will not be repeated here.

[0179] S903. Based on the storage coordinates of the target storage location in the list of devices to be retrieved, calculate the target path from the access control location to the target storage location, and sequentially light up the display units of each device storage location on the target path according to the timing delay to form dynamic light guidance.

[0180] It should be understood that this step, based on a two-dimensional electronic map of the warehouse, uses a shortest path algorithm to calculate the optimal path from the access control to the target storage location. The path is then decomposed into a sequence of nodes, and the display units along the path are illuminated sequentially at preset time intervals. This allows users to quickly and accurately reach the target location along the light strip, realizing a transformation from "static location query" to "dynamic path navigation." For a detailed description of path calculation and dynamic lighting guidance in the above system embodiments, please refer to the above examples; it will not be repeated here.

[0181] S904. Compare the real-time tag code read by the reading device of the target storage location with the target device code in the list of devices to be retrieved, and control the electronic lock of the target storage location to perform device release operation or trigger an abnormal prompt based on the comparison result.

[0182] It should be understood that this step involves real-time acquisition of the tag code of the device being retrieved and performing binary matching. If the comparison matches, the electronic lock is activated and the retrieval is recorded as successful. If the comparison does not match, the feature similarity between the mistakenly retrieved device and the target device is calculated, and the alarm level is dynamically determined. This achieves a closed-loop verification process for retrieval, moving from a "one-size-fits-all alarm" to a "risk-based tiered response." For a detailed description of retrieval verification and tiered alarms in the above system embodiments, please refer to the above examples; it will not be repeated here.

[0183] Based on S901-S904, the warehouse management method provided in this application sequentially links four stages: access control, biometric verification, dynamic lighting guidance, and closed-loop verification for retrieval, forming a fully intelligent closed-loop management system from "application approval—identity verification—path guidance—retrieval verification." This method improves the efficiency and accuracy of equipment retrieval while resisting forgery attacks through composite biometric verification and accurately responding to mis-retrieval events of varying risks through a tiered alarm mechanism, comprehensively enhancing the security and reliability of warehouse management.

[0184] It should be noted that the warehouse mentioned above can be a regular warehouse, the warehouse described in the above scenario, or a skid-mounted warehouse, etc., without specific restrictions.

[0185] Figure 10 A schematic diagram of the structure of an electronic device provided in this application, such as... Figure 10 As shown, the electronic device 1000 includes a processor 1001 and a memory 1002; the memory 1002 stores instructions executable by the processor 1001; when the processor 1001 is configured to execute the instructions, the electronic device implements the method described in the foregoing method embodiments.

[0186] This application also provides a computer-readable storage medium storing computer program instructions thereon; when the computer program instructions are executed by a computer, the computer causes the computer to implement the methods described in the foregoing embodiments. The computer may be an electronic device, a network device, or a manager. The computer-readable storage medium may be a non-transitory computer-readable storage medium, such as a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device.

[0187] This application also provides a computer program product that, when run on a computer, causes the computer to execute the relevant method steps described in the above method embodiments.

[0188] The electronic devices, computer-readable storage media, or computer program products provided in this application are all used to perform the corresponding methods provided above. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods provided above, and will not be repeated here.

[0189] Through the above description of the implementation methods, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0190] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another apparatus, or some features may be ignored or not executed. Furthermore, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0191] The units described as separate components may or may not be physically separate. A component shown as a unit can be one or more physical units; that is, it can be located in one place or distributed in multiple different locations. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0192] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0193] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, essentially, or the parts that contribute to the prior art, or all or part of the technical solutions, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0194] In the description of the embodiments of this application, specific features, structures, materials or characteristics may be combined in any suitable manner in one or more embodiments or examples.

[0195] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A warehouse management system, characterized in that, The warehouse management system includes: Access control devices are installed at the warehouse entrance to collect users' facial images; Multiple device storage slots are located inside the warehouse. Each device storage slot is used to store devices to be retrieved. Each device storage slot includes a display unit, a reading device, and an electronic lock. The display unit is used to display guidance information; the reading device is used to read the tag code of the stored device; and the electronic lock is used to lock or release the stored devices to be retrieved. The central control platform is connected to the access control device and the multiple device storage locations, and is configured as follows: In response to a device access request submitted by a user terminal, an access authorization instruction and a list of devices to be accessed are generated when the access authorization result is approved. Based on the collected user facial image and the access authorization command, biometric verification is performed, and the access control device is controlled to open the access control after the biometric verification result is passed; Based on the storage coordinates of the target storage location in the list of devices to be retrieved, the target path from the access control location to the target storage location is calculated, and the display units of each device storage location on the target path are lit up sequentially according to the time delay to form dynamic light guidance. The real-time tag code read by the reading device of the target storage location is compared with the target device code in the list of devices to be retrieved. Based on the comparison result, the electronic lock of the target storage location is controlled to perform a device release operation or trigger an abnormal prompt.

2. The warehouse management system according to claim 1, characterized in that, The central control platform is configured to compare the real-time tag code read by the reading device of the target storage location with the target device code in the list of devices to be retrieved, and based on the comparison result, control the target storage location to perform a device release operation or trigger an exception prompt, including: If the comparison result between the real-time tag code and the target device code is consistent, control the target storage location to open and allow access, and record the successful retrieval event of the device; If the comparison results between the real-time tag code and the target device code are inconsistent, the alarm level is determined based on the feature similarity between the device in the target storage location and the target device corresponding to the target device code; based on the alarm level, the alarm volume and the alarm information push range are controlled.

3. The warehouse management system according to claim 1, characterized in that, The central control platform is configured to calculate the target path from the access control location to the target storage location based on the storage coordinates of the target storage location in the list of devices to be retrieved, including: Based on a two-dimensional electronic map of the warehouse, the shortest walking path is determined by using the access control location as the starting node and the target storage location as the ending node. The two-dimensional electronic map of the warehouse is a pre-stored floor plan of the warehouse interior in the central control platform. The floor plan is marked with the physical coordinates of each equipment storage location and the walking distance between each adjacent node. The preset path algorithm is used to search the two-dimensional electronic map from the starting node step by step to the ending node, and select the path with the shortest total walking distance. The target path is decomposed into a sequence of nodes, and the display units corresponding to the device storage positions of each node are lit sequentially from the starting node to the ending node according to a preset time interval.

4. The warehouse management system according to claim 1, characterized in that, The central control platform is configured to perform biometric verification based on the collected user facial image and the access authorization command, and to control the access control device to open the access control system after the biometric verification result is successful, including: The facial feature vector of the collected face image is extracted and compared with the standard feature template pre-stored in the access authorization instruction to determine the comprehensive judgment score; Liveness detection is performed based on user action data to determine the liveness detection result; If the comprehensive judgment score is greater than the preset threshold and the liveness detection result meets the preset conditions, the biometric verification result is determined to be passed, and the access control device is controlled to open the access control.

5. The warehouse management system according to claim 4, characterized in that, The central control platform is configured to extract facial feature vectors from the acquired face images, compare them with pre-stored standard feature templates in the access authorization command, and determine a comprehensive judgment score, including: The first similarity is determined based on the Euclidean distance between the facial feature vector and the standard feature template; The second similarity is determined based on the cosine similarity between the facial feature vector and the standard feature template; A comprehensive judgment score is determined based on the first similarity, the second similarity, and the preset maximum distance threshold.

6. The warehouse management system according to claim 1, characterized in that, The central control platform is configured to respond to device access requests submitted by user terminals, perform permission verification, and generate access authorization instructions and a list of devices to be accessed when the permission verification result is approved, including: The device access request is parsed; the device access request includes: the requesting user's identity identifier and the target device number; The administrator privilege database is accessed, and if the user is deemed qualified to apply based on the applicant's identity identifier, the application information is sent to the administrator terminal for approval and confirmation. If the approval result returned by the administrator terminal is approved, the real-time status of the equipment inventory database is queried, available equipment is filtered out to generate the list of equipment to be picked up, and the access authorization instruction containing the user's biometric index and the storage coordinates of the target storage location is generated.

7. The warehouse management system according to claim 1, characterized in that, The warehouse management system further includes an environmental monitoring and self-adjustment module, which includes a temperature and humidity sensor and an adjustment device; the adjustment device is a device for adjusting the temperature and humidity parameters inside the warehouse; the temperature and humidity sensor is used to collect the temperature and humidity signals of the warehouse at preset time intervals and convert them into current environmental values. The environmental monitoring and self-adjustment module is configured to: when the current environmental value is greater than or equal to a preset standard value, determine the required operating power and working time of the adjustment device based on the difference and rate of change between the current environmental value and the preset standard value, and issue an adjustment command to the adjustment device.

8. The warehouse management system according to claim 1, characterized in that, The central control platform is also configured to: Construct a time series of equipment consumption based on historical inventory data; The device consumption time series is input into a preset neural network model for multi-step rolling prediction, and the predicted inventory value at multiple future times is output. When at least two of the predicted inventory values ​​at multiple future time points are lower than the corresponding preset safety stock threshold, a replenishment warning signal is generated and pushed to the administrator terminal.

9. The warehouse management system according to claim 1, characterized in that, The central control platform is also configured to: The system collects multi-source sensor data in real time during the operation of the warehouse management system. The multi-source sensor data includes at least the access control switch status, warehouse temperature and humidity, equipment vibration signals, and network communication traffic. The multi-source sensor data collected in real time is input into a preset anomaly detection model to determine a comprehensive anomaly score; When the overall anomaly score exceeds the preset security baseline, a potential intrusion or equipment failure is identified, and a data circuit breaker operation is executed. The data circuit breaker operation includes: cutting off external data connections, locking all equipment storage locations, and switching the warehouse management system to a local offline operation mode. The local offline operation mode is used to indicate that only the highest-privilege administrator is allowed to perform system recovery operations through a physical key or a local dedicated interface.

10. A warehouse management method, characterized in that, The warehouse management method includes: In response to a device access request submitted by a user terminal, an access authorization instruction and a list of devices to be accessed are generated when the access authorization result is approved. Based on the collected user facial image and the access authorization command, biometric verification is performed, and the access control device is controlled to open the access control after the biometric verification result is passed; Based on the storage coordinates of the target storage location in the list of devices to be retrieved, the target path from the access control location to the target storage location is calculated, and the display units of each device storage location on the target path are lit up sequentially according to the time delay to form dynamic light guidance. The real-time tag code read by the reading device of the target storage location is compared with the target device code in the list of devices to be retrieved. Based on the comparison result, the electronic lock of the target storage location is controlled to perform a device release operation or trigger an abnormal prompt.

Images